What is "vertically partitioned" data, and why is it a natural fit for MPC?

Vertically partitioned data means different organizations hold different attributes (columns) about the same individuals — MPC lets them analyze the joined dataset without ever merging or exposing their columns.

Contrast the two ways data can be split:

• Horizontally partitioned: each party has the same columns for different people (e.g. two hospitals, different patients).
• Vertically partitioned: each party has different columns for the same people (e.g. a bank knows your spending, a hospital knows your diagnoses).

MPC shines for the vertical case: you want to correlate attributes that live in separate silos, but neither side is allowed (legally or competitively) to hand over its raw column. MPC computes the joint result as if the table were merged — without merging it.